8 Best Practices to Generate Code Using AI Tools

welcome devs today we're going to go

over the eight best practices to

generate code using AI tools I'm going

to use three different examples and I'm

going to use three different AI plugins

codium 8 co-pilot and AI assistant let's

begin number one we're going to break

the request into smaller units of work

AI coding assistants perform best when

handling Focus specific tasks number two

provide context in each ask context

helps AI assistants understand the scope

and purpose of the task number three be

clear and specific AI models perform

better with clear and Specific

Instructions number four keep requests

distinct and focused focused asks reduce

confusion and ensure the assistant does

not mix up tasks let's see these four in

action okay so I'm going to use co-pilot

for this example and this is a small

unit of work we're going to verify

credentials we're being distinct on

saying return success or a failure and

we're being very focused on saying use

bcrypt so we're not just letting the AI

think of how will it do this verifying

credentials we're giving it hints or

telling it this information I'm also

going to give it a context of the user

store to make sure that it knows that it

needs to be used in this

answer and now it comes back with our

answer and or our example code and we

see that and we like it so it's up to

you you're still in the driver's seat I

say that that looks pretty good and so

I'm going to import that in now how do I

know if this is exactly right who's your

first consumer of this code we don't

want to check this in and then say well

let's try to use this this in another

piece that's where our unit tests come

in

so this is the whole key to this is that

your your writing of code and your flow

is now just being enhanced by AI we were

able to get some of the example we were

looking at it we might need to update it

or not and then you get your test so now

this looks like a pretty good test to me

and I'm going to go ahead and just for

our example I'm going to go and copy

this and I'm going to paste it in there

now you can see it gave me the whole

class I could have saved that in there

but for the sake of time

I'm just going to run this and have it

in that file and you can see that we

have our makito going and we have our

behaviors tested number five and six

iterate and refine and leverage previous

conversations and generate code these

two work together where AI generated

code might need refinement iterating

helps improve accuracy by building on

previous interactions if you maintain a

coherent flow let's take a look

so I'm going to use AI assistant to

demonstrate the iterate and refine and

use the already existing generated code

so here is our generated code that we've

did and we're going to rightclick on

that and say find problems now ai

assistant is going to go ahead and spit

out a whole bunch of different

descriptions for us and some changes

that is suggesting so it's saying look

you can check for username and passwords

null and stored information there and it

also has come up with some more

information so here is the final revised

method so you'll see it's iterating over

and saying oh here's some stuff

recommended here's a little bit more and

now here's the final part for the

verified credentials so I have that

verified credentials highlighted and I

can simply go ahead and put that in

there and we've iterated over this so

right now that was the problem that it

found

we can go ahead now and test this and we

know that currently it's going to pass

because we're not checking for any nulls

so in the iterating over this we've seen

that well we probably should check for a

username and password being null and not

even go ahead and do that call up or

check for it so let's go ahead now with

this we can actually create some more

tests so I'm going to go use the AI

assist assistant this time time and

generate unit tests so this open up my

unit test and it will start doing this

diff and generate some tests for us so

as this is

generating we'll go ahead and we're just

waiting

here takes a little bit now you're going

to have to read this and understand it

so you need to get there isn't any kind

of like explanation what behaviors

you're testing so once this is done I

can say accept all and then I will need

to go through this and kind of

understand it and sometimes it's a

little bit off but all you have to do is

kind of clean it up you're still the

driver and the driver seat here to make

sure your code's working so let's go

ahead and test this

out and we see that we've met all the

requirements let's see if we have a null

in there so there's a password null so

that we should be getting a um a

verified of a false and then the

username is false too so it's come up

and um put in some uh good mocking and

even an invalid password or wrong

password check so that is the iterating

and refining and making it just that

much better of a codebase now checking

for the username and

password number seven use Advanced

predefined commands for specific

specific asks many AI coding assistants

offer Advanced commands or settings that

trigger more precise and tailored

responses and number eight ask the

explanation when needed understanding

the generated code enhances your ability

to modify or expand it let's dive deeper

into these points with an example so I'm

going to go ahead and check this whole

method and I'm going to do a prepared

command to say

improve and what this will do is this is

a predefined command one of our best

practices to do this and Cody M's going

to go out and come up with some uh

different suggestions so we can extract

credentials we can use this constant

time and this is the timing attacks that

can happen so that's pretty good for the

security let's go ahead and look at it

and it's using the message digest to do

this so let's go head and select that

one and prepare the code to for the

change and we can show the diff we can

see that we're changing this piece right

here and I'm going to go ahead and apply

that um I have to import it so I import

the class and now we have that improved

now um this will tell you a little bit

more about that and we can do some more

um commands here too so I can you know

select let's just select all of this and

and this is the last best practice which

is the explain so I might want or I'm

new here to looking at this

authorization service and I want to have

a better understanding of what this

authorization service is doing and so

this is going to explain the code for me

and explain everything that we've been

doing here so it says it hashes the

provided password and Compares it to

store to Hash using the is equal um

returns true or false on this um that

checks the username if it exists in the

store all of these different flows that

it's doing and its outputs and um you

can see the references that it's doing

so it actually uh referenced the user

store and the authorization service and

is just giving us some examples of how

to use this too if you wanted to use it

so that's the power of these

predefined commands and also to actually

explain your code base to get a better

understanding before you go about doing

your incremental changes or your

requests to wrap up we've explored some

key best practices for using AI tools to

generate code the main takeaway from

this demo is that code generation has

become a standard capability but the

real focus should be on producing

highquality maintainable code with

strong in Integrity this isn't just

about generating code it's about using

AI throughout the entire development

life cycle to assist you while ensuring

you stay in the driver's seat